Post on 25-Sep-2020
transcript
Proposals for increasing benchmarking data quality
of projects measured in COSMIC
Harold van Heeringen, Metrics Consultant Sogeti
Luca Santillo, Agile Metrics (.it)
Friday May 30, Milan, Italy
Outline
• COSMIC characteristics• Comparison with FPA• Benchmarking - challenge• COSMIC Benchmarking committee• Ways to improve benchmarking
in COSMIC
COSMIC
• Second generation FSM • Exists since end 90’s• ISO standard!• More widely applicable than IFPUG
>Business application software domain>Realtime software domain>Infrastructure software domain
>Modern development methods, like RUP>Modern architectures, like SOA>Possible to size reusable components
COSMIC Organisation
• Board>Legal representatives for Canadian law
• International Advisory Committee>Country representatives
• Measurement Practices Committee >Maintenance and improvement of method
• Benchmarking Committee>Harold van Heeringen>Luca Santillo
The road to version 3.0
x COSMIC founded
1998 2000 2002 2004 2006 2008
x ‘Field Trial’ COSMIC-FFP v2.0x v2.1 – ISO 14143/1 conformant
x COSMIC-FFP ISO 19761x v2.2 – ISO 19761 conformant
x COSMIC v3.0
Development
Refinement
COSMIC release 3.0
• Released september 2007• Method is now stable and mature• Many (large) organizations have
adopted the method>Internal estimating & performance
measurement processes work fine>But how about benchmarking processes?
FPA in a nutshell
• Applicability of FPA is strongly dependent on the existence of a data model
• FPA grants function points to both data functions and to logical transactions
• The maximum size of data functions and logical transactions is limited> ILF: 7,10 or 15 FP> EIF: 5, 7 or 10 FP> EI: 3,4 or 6 FP> EQ: 3,4 or 6 FP> EO: 4,5 or 7 FP
COSMIC in a nutshell
• COSMIC is not dependable on the existence of a data model
• COSMIC values data movements within functional processes and does not explicitlyreward data functions
• This value is not limited per functional process>Theoretically the size of a COSMIC functional
process can be any number between 2 and infinity
Most important differences
FPA COSMIC
Applicable on Domain Business Software Business, Real-time, Infrastructure Software
Data model required? Required Not required (but useful)Measurement of separate components?
Not possible Possible
Size limit per function Yes Size is not limited
Benchmarking data Many (ISBSG R10: n=3108)
Few (ISBSG R10: n=110)
Measurement of processing functionality
No No, but local extensions are possible
Early sizing Based on data model Based on process model
Possible reasons to adopt COSMIC
• New development methods new functional documentation standards FPA less applicable
• The organization wishes to measure software in real-time, telecommunications or infrastructure domains
• The organization wishes to measure the size of reusable components
• The organization wishes to measure more accurately. A very complex function should get more points than a complex function
• The organization is organized in a way that a department develops only part of the user functions and estimations have to be made at that department level.
Possible reasons not to adopt COSMIC
• Loads of experience data in FPA• Organizations develop in 3GL
languages in ‘old fashioned’environments
• Lack of benchmarking data !
Benchmarking challenge
• ISBSG database R10>Total projects: 4.106>IFPUG projects: 3.108>COSMIC projects: 110
• Benchmarking challenge: Increase available benchmarking data in COSMIC
• COSMIC Benchmarking Committee
Benchmarking Committee – Goals
• Goal: To increase the amount of benchmarking data available to the industry
• Initiatives>Conversion formulae>Increase adoption>Lower ISBSG data submission hurdle>Ask for data
Initiatives – Conversion formulae
• Method>Double measurements of a number of projects
by experienced analysts in both methods
• Projects>26 projects for a variety of clients>New developments>Business applications>Variety of branches>Early requirements>Often low quality documentation
Initiatives – Conversion formulae
N=26Y(CFP) = 1,22 (FP NESMA) – 64R2 = 0.97
y = 1,22x - 64R2 = 0,97
0200400600800
100012001400160018002000
0 500 1000 1500 2000
# C
FP --
>
# FP -->
Results
• Use one formula for the conversion> CFP = 1.22 FP (Nesma) – 64> FP (Nesma) = 0.82 (CFP) + 52> Projects > 200 FP
• Advising clients: Use Transition framework
Input
Development tools Java
Distribution of work Onshore OffshoreConstruction 0% 100%System test 0% 100%
System test strategy TMap Medium
Delivery test Yes
Complexity Medium
Size 1235 FP
Start date 01-08-07
Input
Development tools Java
Distribution of work Onshore OffshoreConstruction 0% 100%System test 0% 100%
System test strategy TMap Medium
Delivery test Yes
Complexity Medium
Size 1500 COSMIC =1282 FP
Start date 01-08-07
Duration in weeks 15 17 19 21 23 25 27 Delivery for acceptance 14-11-07 28-11-07 12-12-07 26-12-07 9-01-08 23-01-08 6-02-08 Total effort 9462 8106 7026 6342 5916 5646 5400 Effort per FP 15,77 13,51 11,71 10,57 9,86 9,41 9 Totaal cost € 413.352 € 354.126 € 306.936 € 277.056 € 258.414 € 246.594 € 236.022 Cost per FP € 689 € 590 € 512 € 462 € 431 € 411 € 393
Initiatives – Increase adoption
• Translate COSMIC documents>Spanish>Italian >Dutch>Japanese>Chinese>Turkish>French
Initiatives – Increase adoption
• White papers>Almost 100 available>www.lrgl.uqam.ca/cosmic-ffp
• Topics:>COSMIC + data warehouse systems>COSMIC + SOA architectures>COSMIC + Real-time systems>COSMIC + enhancement / maintenance>etcetera
Initiatives – Lower ISBSG submission hurdle
• ISBSG>Submit data and receive a benchmark report
• Current ISBSG COSMIC questionnaire>138 questions
• Initiatives>Restructuring questionnaire: A / B / C>Webclient project submission>Export facility in tools>(Convert ISBSG database)